A Combined Data Analytics and Network Science Approach for Smart Real Estate Investment: Towards Affordable Housing

Abstract

Sophisticated tools for smart management and public services are crucial aspects of smart cities and especially affordable housing. In this context, a novel algorithm is introduced, which assists a user to identify locations for real estate investment. The methodology involves an application of data analytics for selection of top attributes of real estate for a user, and based on these attributes stacks of machine learning algorithms like decision trees, principal component analysis (PCA), and K-means clustering identify the location for investment. While data analytics comprising statistical modeling and machine learning techniques can compute the important attributes and thereby identify locations, it is nontrivial to get good insight at the scale of a large complex network consisting of hundreds of attributes and locations. This is mainly due to the underlying assumptions of i.i.d (independent and identically distributed) on random variables of many learning algorithms. Network science provides the necessary tools to analyze interactions and relations among entities in large networks considering the interdependencies of variables. In this chapter, a network created from the locations outputted by machine learning layers is described that utilizes network measures like eigen centrality that helps a user to determine the best location for investment, while providing deeper insight into the location identification problem. In addition, simulation of network dynamics provides the most influential and stable attribute of the designed real estate complex network, in the presence of the random link weight perturbations. Real estate investment comprises many attributes that can be categorized into social, economic, governmental, and environmental. Of all these, only real estate factors are considered in this work. However, the same work can be extended to other factors as well.